1. Technical Field the Disclosure
The present disclosure relates generally to an electronic device, and more particularly to a configuration of an electronic device for voice recognition.
2. Description of the Related Art
In general, people unique voices and the voice itself may be used to authenticate a person.
For example, a voice recognition device may recognize the voice of a particular person by using a voice recognition model having collected the voice of the particular person and information on the voice, which is commonly referred to as “speaker verification”. A voice recognition device may also distinguish the voice of a speaker by using a voice recognition model trained in advance for recognizing voices of a plurality of people, which is commonly referred to as “speaker identification”.
A voice recognition device using the speaker verification or the speaker identification may train the voice recognition model by using a particular phrase and, in this case, recognize a voice only when a particular speaker speaks the particular phrase, thereby having a higher security performance.
Further, a voice recognition device may recognize a voice by using an isolated word recognition method that recognizes only a predetermined particular word. The isolated word recognition method generates a template of each particular word and compares it with an input speech. The voice recognition device using the isolated word recognition method recognizes only a predetermined particular word, and thus, has a relatively high voice recognition rate and has a relatively low voice recognition failure rate due to background noise. Accordingly, the isolated word recognition method has fewer calculations and uses less memory as compared to Large Vocabulary Speech Recognition (LVSR), in which all speech can be converted into text, and natural voice recognition. Therefore, the isolated word recognition method can be easily used in a portable terminal device.
However, a conventional voice recognition device has a problem performing a voice recognition method having a large calculation amount, a low power processor is mounted therein.
Further, a conventional voice recognition device has a disadvantage of large power consumption when a high performance processor is mounted therein, because high performance pre-processing and high performance voice recognition are performed.